AI Security: Enterprises Facing Breaches and Adversarial Attacks, But Budgets on the Rise
A recent survey of 150 IT security and data science leaders has revealed some concerning trends when it comes to the security of artificial intelligence (AI) models within enterprises. The survey found that a staggering 98 percent of companies consider AI models to be essential to their business operations. However, a worrying 77 percent of these companies reported experiencing breaches in their AI systems within the last year.
Despite this alarming statistic, the survey also found that only 14 percent of IT leaders claim their companies are actively planning and testing for adversarial attacks on their AI models. This lack of proactive preparation is concerning given the growing number of AI models being deployed by companies. On average, businesses currently have 1,689 AI models in production.
Recognizing the urgent need to address these vulnerabilities, the survey found that 94 percent of IT leaders are allocating budgets for AI security in 2024. However, the confidence in these security budgets is moderate, with only 61 percent expressing high confidence in their effectiveness. Furthermore, a significant 92 percent of companies are still in the process of formulating a comprehensive security plan for their AI systems.
The concerns over AI security are not without merit. Chris Sestito, co-founder and CEO of HiddenLayer, has described AI as “the most vulnerable technology ever to be deployed in production systems.” This vulnerability is further exacerbated by the presence of shadow IT and third-party AI integrations. The survey found that 61 percent of companies acknowledged shadow IT as an issue, with a staggering 89 percent expressing concerns about vulnerabilities in third-party AI integrations. Additionally, 75 percent of respondents viewed these integrations as posing a higher risk compared to other security threats.
To enhance AI security, the survey offered several recommendations. Firstly, it is crucial for companies to have a clear understanding of how AI is being used within their organization. This understanding will enable them to better identify potential risks. Additionally, continuous anomaly detection and regular assessments of AI models’ robustness against attacks are crucial. By implementing these measures, companies can significantly improve the security of their AI systems.
As the deployment of AI continues to accelerate, it is imperative for enterprises to prioritize AI security. The survey’s findings highlight the urgent need for companies to invest in robust security measures and proactive planning to safeguard their AI models and mitigate the risk of breaches and adversarial attacks.
Analyst comment
Neutral news.
As an analyst, it is expected that the market for AI security will see growth in the coming years. The increasing number of AI models being deployed and the high incidence of breaches indicate a need for companies to invest in robust security measures. With budgets being allocated for AI security and recommendations provided, companies have an opportunity to improve the security of their AI systems, mitigating the risk of breaches and adversarial attacks.
